Auditees case-selection model for evaluating taxpayer corporate tax compliance in Kenya

Abstract

Tax compliance rate in Kenya is estimated to be approximately below 65%. It is important for the government to place measures that ensure improved tax compliance rate comparable with benchmark countries like Sweden, whose tax compliance rate stand at 93%. One measure implemented in Kenya Revenue Authority has been to conduct scrutiny assessments on the taxpayer fraternity. However, success in scrutiny assessments in addressing payment and reporting compliance is largely dependent on the cases selected for audit. A major challenge has been in the possibility of selecting of an honest taxpayer and failure to take up the potential under-reporter, scenarios which are both costly to the tax administration. Whereas the honest taxpayer will feel unfairly selected for scrutiny, under-reporters escape the purview of the authority. This study presents a data mining based approach aimed at addressing the case-selection challenge. A classification model built using historical taxpayer audit data and decision tree algorithm was used to predict the compliance status of taxpayers in a case-selection application prototype. Experimental results using limited taxpayer data for the period year 2014/2015 indicate that the model is effective and fit for case-selection with an accuracy rate of 65% and prediction efficiency of 65% in identifying non-compliant taxpayers. Moreover, with more sources of taxpayer information and increased quantity of data, the accuracy and prediction efficiency is expected to improve significantly. It is recommended that Kenya Revenue Authority adopts this approach to improve the traditional case-selection by auditors‟ for corporate tax as well as other tax obligations such as Individual income tax, VAT, and custom duties administered by the Kenyan government.